lisa a. cavanaugh, james r. bettman, and mary …...aquino, and levy 2007; winterich, mittal, and...

Journal of Marketing Research, Ahead of PrintDOI: 10.1509/jmr.10.02191

© 2015, American Marketing AssociationISSN: 0022-2437 (print), 1547-7193 (electronic)

*Lisa A. Cavanaugh is Assistant Professor, Department of Marketing,Marshall School of Business, University of Southern California (e-mail:lisa. [email protected]). James R. Bettman is Burlington Industries Professor of Business Administration, Fuqua School of Business, DukeUniversity (e-mail: [email protected]). Mary Frances Luce is Senior Associ-ate Dean for Faculty and Robert A. Ingram Professor, Fuqua School ofBusiness, Duke University (e-mail: [email protected]). This article is basedon the first author’s dissertation. The authors thank Barb Fredrickson andthe PEP lab for their encouragement and support of this program ofresearch, Kristin Diehl and Debbie MacInnis for their helpful comments,and Nicholas J. Jackson for his statistical assistance. Jeffrey Inman servedas associate editor for this article.

LISA A. CAVANAUGH, JAMES R. BETTMAN, and MARY FRANCES LUCE*

Marketers often employ a variety of positive emotions to encourageconsumption or promote a particular behavior (e.g., buying, donating,recycling) to benefit an organization or cause. The authors show thatspecific positive emotions do not universally increase prosocial behaviorbut, rather, encourage different types of prosocial behavior. Four studiesshow that whereas positive emotions (i.e., love, hope, pride, andcompassion) all induce prosocial behavior toward close entities (relativeto a neutral emotional state), only love induces prosocial behavior towarddistant others and international organizations. Love’s effect is driven by adistinct form of broadening, characterized by extending feelings of socialconnection and the boundary of caring to be more inclusive of othersregardless of relatedness. Love—as a trait and a momentary emotion—is unique among positive emotions in fostering connectedness that otherpositive emotions (hope and pride) do not and broadening behavior in away that other connected emotions (compassion) do not. This researchcontributes to the broaden-and-build theory of positive emotion bydemonstrating a distinct type of broadening for love and adds animportant qualification to the general finding that positive emotionsuniformly encourage prosocial behavior.

Keywords: positive emotions, prosocial behavior, love, social connection,broaden and build

Online Supplement: http://dx.doi.org/10.1509/jmr.10.0219

Feeling Love and Doing More for DistantOthers: Specific Positive EmotionsDifferentially Affect Prosocial Consumption

Prosocial behavior is of great interest to consumers andmarketers alike (e.g., Agrawal, Menon, and Aaker 2007).Behaviors such as civic participation, volunteering, donat-ing money, or buying products that benefit a good cause are

often regarded as undifferentiated (Collett and Morrissey2007). However, most organizations that promote prosocialbehaviors desire a very specific consumption behavior (theywant people to, e.g., buy, recycle, donate, or vote in a par-ticular way) as opposed to just any prosocial or helpfulbehavior. Thus, understanding when and why people engagein specific prosocial consumption behaviors is of greatinterest to consumer behavior researchers, sociologists, psy-chologists, and practitioners (e.g., Batson et al. 2008; Pili-avin and Charng 1990).Although a variety of personal, motivational, or contex-

tual factors might induce prosocial behavior (Batson et al.2008), organizations often rely on positive emotions in theirmarketing and advertising to encourage such behaviors.Consumer products companies (e.g., General Electric, Nike,Procter & Gamble), nonprofit organizations (e.g., American

Red Cross, The Nature Conservancy), and even politicalcandidates (e.g., the Obama presidential campaign) regularlyemploy positive, but often diffuse, emotional themes in adver-tising. In the prosocial domain, marketers often use positiveemotions interchangeably. The underlying assumption seemsto be that all positive emotions increase all prosocial behaviors(i.e., if consumers feel good, they are more likely to do good).Indeed, previous researchers have linked generalized posi-tive affect or the effects of positive versus negative affect tomultiple prosocial behaviors, including helping, generosity,interpersonal understanding, and monetary donations (e.g.,Small and Verrochi 2009; for a review, see Isen 2001).However, the effects of different specific positive emotionshave not generally been considered for prosocial behavior(for an exception contrasting amusement and gratitude, seeBartlett and DeSteno 2006) or charitable giving.We examine the general question of whether specific

positive emotions differentially motivate particular behav-iors by testing the novel hypothesis that specific positiveemotions may have different effects on prosocial behaviordirected toward close versus distant others. Conditions ofchronic poverty and natural disasters (e.g., famine, floods,earthquakes) in many of the poorest areas of the world (e.g.,sub-Saharan Africa, South Asia; United Nations 2011) oftenprompt government and nonprofit agencies in those areas tosearch for help from people abroad (e.g., the United States).In addition, organizations regularly solicit donations to pre-empt and address major global issues (e.g., deforestation,illiteracy, disease) across continents. Thus, consumers areoften asked to contribute to distant others about whom theyhave no personal knowledge and to organizations address-ing problems with which they have no personal experience.Aggregate giving data suggest that these requests tend to beat a considerable disadvantage compared with those fromcloser organizations that are known entities. For example,U.S. citizens gave nearly $316 billion to charitable organi-zations in 2013, the majority of which went to local reli-gious (32%) and local educational (13%) organizations,with only 6% of all giving going to international organiza-tions and international disasters (Giving USA 2013). We usethis important and challenging problem of promoting givingto distant others as a context within which to study our pro-posed approach of using specific positive emotions to pre-dict and influence behavior.In marketing, research has shown that love, hope, pride,

and compassion have important influences on consumers(e.g., Belk and Coon 1993; Cavanaugh et al. 2011; Mac -Innis and De Mello 2005; Small and Verrochi 2009), and allare regularly employed in marketing appeals in prosocialconsumption and charitable giving contexts. How might theuse of different positive emotions affect the success ofappeals for helping distant others? We hypothesize andshow that whereas positive emotions (vs. neutral emotionalstates) typically enhance prosocial behavior aimed at closeothers, only love (not hope, pride, or compassion) enhancesprosocial behaviors aimed at distant others. Although love,hope, and pride share positive feelings, love is distinct fromhope and pride in that it also generates feelings of socialconnection, enhancing consumers’ propensity to feel caringand exhibit concern toward those with whom they are notrelated (i.e., others with whom psychological and physical

proximity are not shared). Thus, love ultimately changes theboundary of caring and concern to include more distant oth-ers. We also examine whether social connection alone issufficient for giving to distant others by comparing lovewith a closely related emotion, compassion, which alsoenhances social connection; however, compassion does sowhile producing both positive and negative feelings.Notably, love, but not compassion, increases giving to dis-tant others, validating our contention that the combinationof social connection and positive feelings (compared withthe co-occurrence of positive and negative feelings found incompassion) generates a specific form of broadening asso-ciated with prosocial behaviors toward distant others.Thus, our research contributes to the consumer emotions

literature, the broaden-and-build theory of positive emotion(Fredrickson 1998, 2001; Fredrickson et al. 2008), and theunderstanding of prosocial behavior. We enrich the con-sumer emotions literature by expanding the set of positiveemotions and mechanisms (e.g., broadening) considered.With respect to the broaden-and-build theory, previous testshave shown that all positive emotions broaden in a similarway, leading to a wider range of attention, thoughts, andactions. However, as we have noted, our research is the firstto suggest and show that love broadens in a particular way,by shifting the boundary of caring and sense of social con-nectedness toward distant others. The finding that a specificpositive emotion broadens in a distinct way, leading to pre-dictable outcomes that are differentiable from other positiveemotions, is an important contribution to the broaden-and-build theory of positive emotion (Fredrickson 1998, 2001,2009; Fredrickson et al. 2008). Documenting that differen-tial forms of broadening are possible also provides newinsight to the consumer emotions literature by providing anew set of characteristics (beyond common notions such asvalence and arousal) with which differences among emo-tions may be conceptualized and tested. Finally, our findingscontribute to the prosocial behavior literature by distinguish-ing different types of beneficiaries of prosocial behavior(close vs. distant others) and by challenging the assumptionthat positive emotions generally and uniformly encourageprosocial behaviors. Again, this insight follows from ourdemonstration that not only does broadening represent amechanism by which positive emotion generates action (asestablished by Frederickson) but, further, different positiveemotions broaden differently (as we establish herein).First, we review the prosocial behavior literature and

identify an important and underexplored dimension ofprosocial behavior: beneficiary focus. We then review theconsumer emotions literature and describe the nature andfunction of specific positive emotions (i.e., love, hope,pride, and compassion), identifying both social connectionand the absence of co-occurring negative feelings as key tolove’s novel effect on behaviors. We then argue that lovecan lead to certain types of prosocial behavior (i.e., behav-iors that benefit distant others). Four studies show thatlove—either as a persistent trait or momentary emotion—isunique among positive emotions in promoting prosocialbehaviors toward distant others. We demonstrate love’seffect by both measuring (for dispositional emotion) andmanipulating emotion (using personal memories and adver-tisements) and showing its impact on multiple distant other

2 JOURNAL OF MARKETING RESEARCH, Ahead of Print

Feeling Love and Doing More for Distant Others 3

beneficiaries (e.g., distant individuals, international humani-tarian and environmental organizations).

PROSOCIAL BEHAVIORPrior research has shown that designated beneficiaries

can influence the likelihood of consumers purchasing prod-ucts and supporting fundraising appeals (Small and Verrochi2009; Strahilevitz and Myers 1998). Consumers are morelikely to donate when fundraising benefits an identifiablevictim (Small, Loewenstein, and Slovic 2007) or generatessympathy and compassion by featuring a sad-faced victim(Small and Verrochi 2009). Individual differences in per-sonal relatedness, prosocial personality characteristics, gen-der identity, and moral identity also influence whether con-sumers help or give (Penner and Finkelstein 1998; Reed,Aquino, and Levy 2007; Winterich, Mittal, and Ross 2009).For example, Winterich, Mittal, and Ross (2009) find thatwomen who reported higher importance of moral identitywere more likely to donate to out-groups. A common featureacross these studies is that they involve some perception ofa designated beneficiary—that is, the people or cause thatwill benefit from the prosocial behavior.One important dimension along which prosocial behaviors

vary is beneficiary focus, and beneficiaries can be describedin terms of distance, broadly interpreted. The bene ficiariesof prosocial behaviors (i.e., the people or entity helped) canvary widely. They can range from psychologically and geo-graphically close others (e.g., local groups, parks, neigh-bors) to more distant others (e.g., international groups, rain-forests, refugees). In general, positive feelings makeconsumers more willing to help close others (Waugh andFredrickson 2006)—that is, people more psychologicallynear to them (e.g., relatives, neighbors, local communitymembers). Such psychological distance to beneficiaries canbe influenced by many things (e.g., geographic distance),not just social identity (e.g., the in-group/out-group distinc-tion studied by Winterich, Mittal, and Ross 2009). Forexample, even among generally unknown or even potentialout-group beneficiaries, psychological distance can be animportant differentiator among classes of appeals (e.g.,domestic vs. foreign aid funds).

POSITIVE EMOTIONSMarketing research on specific emotions has historically

emphasized contrasts between positively and negativelyvalenced emotions (e.g., Chang and Pham 2013; Griskevi-cius et al. 2009) and differences between specific negativeemotions such as anger, fear, sadness, and disgust (e.g., Cry-der et al. 2008; Lerner and Keltner 2001; Lerner, Small, andLoewenstein 2004; Raghunathan and Pham 1999). Positiveemotions often have been characterized as relatively undif-ferentiated (Ellsworth and Smith 1988; Isen 2001; Smithand Ellsworth 1985), with the exception of arousal differ-ences. Consumer and marketing researchers who haveexamined different positive emotional states have over-whelmingly emphasized happiness (Sauter 2010) and com-pared positive emotions characterized by or differinglargely in terms of arousal, such as upbeat versus warm feel-ings (Burke and Edell 1989), excitement versus content-ment (Kim, Park, and Schwarz 2010), pride versus content-ment (Griskevicius, Shiota, and Nowlis 2010), happiness

versus peacefulness (Agrawal, Menon, and Aaker 2007),happiness versus calmness (Labroo and Rucker 2010), andnonrelaxed versus relaxed positive emotion (Pham, Hung,and Gorn 2011). Moreover, researchers have not consideredwhether specific positive emotions could differentiallyinfluence charitable giving.1We go beyond arousal-based distinctions by examining a

set of positive emotions (i.e., love, hope, pride, and compas-sion) that we hypothesize will have specific effects onbehaviors that benefit distant others. Our approach to exam-ining distinct positive emotions is based on the broaden-and-build theory (Fredrickson 1998, 2001), which describesthe nature and general shared function of positive emotionsas distinct from negative emotions. Unlike negative emo-tions, which narrow people’s focus to help manage andrespond to aversive situations, positive emotions function tobroaden attentional, cognitive, and motivational scope toallow for new perspectives and experiences (e.g., Fredrick-son 1998, 2001; Fredrickson and Branigan 2005; for adivergent view, see Gable and Harmon-Jones 2008). Broad-ening is not a function of arousal (Fredrickson and Branigan2005). Our framework both leverages and contributes to thebroaden-and-build theory of positive emotion by using ananalysis of the specific properties of love, hope, pride, andcompassion to derive hypotheses regarding a unique type ofbroadening that we predict will be specific to love. Love,hope, and pride are all positive in valence (Fredrickson1998; Oveis, Horberg, and Keltner 2010) but, we contend,differ in their potential broadening effects.LoveConceptually, the word “love” has been used to capture a

range of feelings involving proximity maintenance. Withinthe marketing and consumer psychology literature streams,“love” has often been used to refer to what are actually themore specific emotions of desire and compassion (Belk andCoon 1993; Goetz, Keltner, and Simon-Thomas 2010;Griskevicius et al 2009; Oveis, Horberg, and Keltner 2010).Although romantic/sexual desire is an interesting topic, it isnot the type of love we examine here. Instead, we focus onthe emotion of love experienced in companionate relation-ships and distinguish that type of love’s effects from thoseof compassion. According to the triangular theory of love,companionate love is characterized by the presence of com-mitment and intimacy without passion (Sternberg 1986) andis distinct from romantic love (passion + intimacy), fatuouslove (passion + commitment), and liking (intimacy; Stern-berg 1986). We define love in terms of feelings of warmthand affection toward platonic others (i.e., family andfriends) in close, nonsexual relationships. Notably, this isthe type of love people most frequently report experiencing,and it is often depicted in marketing appeals (e.g., GeneralMills, Johnson & Johnson, Procter & Gamble).Emotion theorists broadly agree that emotions differ in

themes and serve distinct functions (e.g., Lazarus 1991).Love functions to foster relationships between humanbeings. Love (not desire) influences bonding (Gonzaga etal. 2006) and feelings of warmth and closeness (Fitness and

1Only one previous article (Small and Verrochi 2009) has examined theeffect of specific emotions (happiness vs. sadness) on charitable giving.

Fletcher 1993) in relationships. Notably, research has foundloving-kindness meditation to heighten feelings of connec-tion toward novel people at both explicit and implicit levels(Hutcherson, Seppala, and Gross 2008).HopeHope is described as a person’s “yearning for better and

believing the wished-for improvement is possible” (Lazarus2006, p. 16). Hope signals that a concrete positive goal isexpected, and it reflects a capability to derive pathways todesired goals and to motivate goal pursuit (Snyder et al.1991). We define hope in terms of feelings that an expendi-ture of energy or effort could result in achieving a valuedpositive change in outcome.Hope functions to influence perception of goal obstacles

and to sustain effort (Ellsworth and Smith 1988; Smith andEllsworth 1985; Snyder et al. 1991) toward goals for oneselfand close others (Reichard et al 2013). Hope further func-tions to enhance coping potential and expectations (Mac -Innis and De Mello 2005).PridePride is described as “enhancement of one’s ego-identity by

taking credit for a valued achievement” (Lazarus 2006, p. 16)or experiencing enhancement of one’s self or social worth bybeing credited for a highly valued accomplishment (Lazarus1991). Pride involves internal attributions and self-credit forvalued events (Lazarus 2006) such that a person feels goodabout him- or herself; it is considered a self-conscious emo-tion. We define pride in terms of feelings of personalresponsibility for achieving a valued positive outcome.Pride functions to provide information about a person’s

current level of status in a group (Tracy and Robins 2007).Because pride involves attribution of positive events to theself (Roseman, Antoniou, and Jose 1996), it also is a sociallydisengaging emotion, promoting increased distance betweenthe self and others (Kitayama, Mesquita, and Karasawa 2006).CompassionAn emotion that may be closer to love is compassion.

Some view compassion as a distinct emotion (Lazarus1991), whereas others view it as a variant or blend of loveand sadness (Shaver et al. 1987). Compassion is describedas “the feeling that arises in witnessing another’s sufferingand that motivates a subsequent desire to help” (Goetz,Keltner, and Simon-Thomas 2010, p. 2) and helps explainwhy sad-faced children increase observer giving (Small andVerrochi 2009). Compassion motivates caretaking of weakor suffering others when exposed to another’s harm (Oveis,Horberg, and Keltner 2010). Notably, compassion and lovediffer in terms of antecedent events: whereas love’santecedents are positive, compassion’s antecedents arenegative (Goetz, Keltner, and Simon-Thomas 2010). Owingto compassion’s focus on alleviating evident suffering (i.e.,a negative antecedent) and its co-occurring negative andpositive feelings, we argue that the broadening associatedwith love, which we hypothesize to be the basis for oureffects, should not be evident for compassion.In summary, specific positive emotions have distinct

functions and lead to different levels of social connection.We conjecture that these distinctions cause them to differ in

their potential broadening effects with important implica-tions for behavior. Love is distinct from hope and pride interms of its generation of social connection. Love is alsodistinct from compassion, which can enhance social connec-tion but commingles positive and negative feelings. Next, wepresent our theory and hypotheses of why love has uniquebehavioral effects within the realm of prosocial behavior.

CONCEPTUAL DEVELOPMENT: THEORY ANDHYPOTHESES

H1Previous findings have suggested that when people

experience positive emotions, they help close others, con-sistent with broaden-and-build (e.g., Waugh and Fredrick-son 2006) and related theories. Thus, we hypothesize thatlove, hope, and pride will lead to helping close others.Although this hypothesis is consistent with prior findings onpositive affect and helping (e.g., Isen 2001), we believe it isimportant first to empirically replicate effects consistentwith previous work before presenting our unique contribu-tion. Thus, we show that the positive emotions we examineall affect behaviors toward close others. Our focal contribu-tion is then to show that specific positive emotions actuallylead to different types of prosocial behavior—namely, dif-ferential effects of specific positive emotions on behaviorsthat benefit distant others, as we articulate in H2 and H3. InStudies 1 and 2, we measure behaviors toward close othersin addition to behaviors toward distant others. For theseclose-other behaviors, we expect to replicate previous find-ings related to positive valence. More formally, we hypothe-size the following:H1: Positive emotions increase contributions to close others

relative to a neutral emotional state.H2 and H3In contrast to the uniform predictions for all positive

emotions in H1, we anticipate that only love and not otherpositive emotions (i.e., hope, pride, and compassion) willinduce contributions to distant others. As we outlined previ-ously, love promotes a level of social connection (i.e., feel-ings of closeness and enhanced relationship with others)that hope and pride do not. Specifically, love should widenthe range (in terms of type and number) of cared-for people.This propensity to increase social connection to distant oth-ers distinguishes love from hope and pride; thus, lovebroadens in a way hope and pride do not.We attempt to clarify the role of social connection more

completely by comparing love and compassion. Both loveand compassion are characterized by high levels of socialconnection, but love is characterized by positive feelings,whereas compassion is characterized by co-occurring posi-tive and negative feelings. We hypothesize that positiveemotion–based broadening is a precondition for the effectsspecified, and thus, we do not expect that compassion willhave effects on giving to distant others.These distinctions and the following studies are the first

examination of the possibility of different forms of broaden-ing generated by specific positive emotions. Note that ourprediction that love (but not hope, pride, compassion, or neu-tral emotions) will increase donations to distant others runs



counter to the intuitive notion that love would cause people tofocus resources only on one’s loved ones (e.g., friends andfamily, who are presumably the focal target of love, featuredin advertising). We derive our (opposite) hypothesis by com-bining a specific analysis of the function of love with the“broaden” aspect of broaden-and-build theory. Thus, loveserves to bond people with others, but, consistent with thestatus of love as a clearly positive emotion, we hypothesizethat it actually does so in a way that results in effects on notonly those who are close but also those who are further away.In summary, we expect consumers experiencing love to

increase prosocial behaviors that benefit distant others morethan the other specific positive emotions we examine and aneutral state. More formally, we hypothesize the following:H2: Love (but not hope, pride, or compassion) increases contribu-

tions to distant others relative to a neutral emotional state.H3: Compared with hope and pride, the impact of love on

behaviors that benefit distant (but not close) others is medi-ated by love’s impact on feelings of social connection.

These hypotheses allow for more precise predictionsregarding positive emotion by linking the functions of spe-cific emotions with particular features of behavior. We haveproposed that prosocial behaviors can be characterized interms of beneficiary focus (i.e., close vs. distant others). Wehypothesize that love increases prosocial behaviors thatbenefit distant others (relative to hope, pride, compassion,and neutral emotions) owing to its tendency to increase feel-ings of social connection while coupled with positive feel-ings (but not mixed feelings). Thus, love produces a form ofbroadening not associated with all positive emotions. Weexamine these predictions across four studies using bothmeasured and manipulated emotions.

STUDY 1: DISPOSITIONAL LOVE AND HOPEDIFFERENTIALLY PREDICT PROSOCIAL

CONSUMPTION BEHAVIOR FOR CLOSE ANDDISTANT OTHERS

Study 1 tests whether people’s propensity to experiencecertain positive emotions influences their willingness toengage in prosocial behaviors that benefit close and distantothers. Specifically, Study 1 tests whether dispositional loveand dispositional hope lead to different patterns of prosocialconsumption behavior. We expect prosocial behaviors thatbenefit close others to reveal general (i.e., undifferentiated)effects of positive emotion on helping (i.e., both dispositionallove and dispositional hope should increase behaviors thatbenefit close others), consistent with previous research. How-ever, we expect dispositional love (but not dispositional hope)to predict increased behaviors that benefit distant others.Method and ProcedureParticipants and setup. Eighty-two students participated in

a 20-minute study on feelings and consumer choice. The sam-ple consisted of 37 men, 44 women, and 1 respondent whoomitted gender, ranging in age from 18 to 41 years (M = 21.4years, SD = 3.2). To dissociate the emotion procedure from thedependent measures of interest, participants were told that dif-ferent researchers had pooled together their respective ques-tionnaire packets and that they would be completing three sep-arate studies, which included a filler task. Study 1 used a

measured, within-subject design in which dispositional emo-tions were measured for each participant and social distanceof beneficiary was manipulated within subjects (close/distant).Dispositional emotion measures. Each participant com-

pleted multi-item measures for dispositional love (six items)and dispositional hope (seven items; Shiota 2004; Shiota,Keltner, and John 2006). For each item, participants wereasked to indicate the extent to which each statement accu-rately described them on a seven-point scale (1 = “stronglydisagree,” and 7 = “strongly agree”; for all items, see theWeb Appendix). Dispositional means were standardizedacross participants.Prosocial consumption intentions. Later in the study, ses-

sion participants were asked to complete a paper-and-pencilconsumer choice survey gauging the likelihood of engagingin various consumption behaviors over the coming year on aseven-point scale (1 = “extremely unlikely,” and 7 =“extremely likely”). The list included eight prosocial con-sumption items as well as filler items (e.g., “see a foreignfilm,” “attend a live music concert”). The prosocial consump-tion items were designed to tap two types of beneficiaries—close and distant others. We averaged the four close-othersitems (a = .71; e.g., “donate used items/clothing to a chari-table organization to help local families in need”) to create aclose-others behavior score. We averaged the four distant-other behaviors (a = .80; e.g., “donate money to a charitableorganization benefiting rainforest conservation in foreigncountries”; for all items, see the Web Appendix) to create adistant-others behavior score. Pretest participants (N = 31)had assessed who would benefit from the behavior for eachof these items on a seven-point scale (1 = “close others,”and 7 = “distant others”). The distant-other behavior itemswere perceived to benefit distant others (Mdistant = 5.3) sig-nificantly more than the close-other behavior items (Mclose =2.6; t(30) = 15.82, p < .0001).ResultsPreliminary analyses. Preliminary analyses on the dispo-

sitional emotion measures showed that the measured emo-tion subscales were reliable (love: a = .80; hope: a = .81).Hypothesis tests. We predicted that dispositional love but

not dispositional hope would be associated with prosocialconsumption behaviors that benefit distant others. To test thisprediction, we ran a 2 (dispositional emotion: love, hope) ¥ 2(distance: close, distant) mixed-effects model with subject-random intercept and distance-random slope to account forrepeated measurements within subjects. We observed a sig-nificant effect for dispositional love (F(1, 79) = 6.57, p < .01),dispositional hope (F(1, 79) = 7.09, p < .009), and distance(F(1, 79) = 204.31, p < .0001) and a significant interactionbetween distance and dispositional hope (F(1, 79) = 5.65, p <.02).2 We found no significant differences for filler items.

2Effects are consistent when each behavior type is analyzed separately—that is, when both dispositional love and dispositional hope scores areentered simultaneously into a model for behaviors benefiting distant others(F(2, 79) = 8.85, p < .0003) and for behaviors benefiting close others (F(2,79) = 13.67, p < .0001). Dispositional love (B = .58; F(1, 79) = 14.41, p <.0003), but not dispositional hope (B = –.05; F(1, 79) < 1, n.s.), was a sig-nificant predictor of behaviors benefiting distant others. In contrast, bothlove (B = .35; F(1, 79) = 6.55, p < .01) and hope (B = .33; F(1, 79) = 7.08,p < .009) predicted behaviors that benefit close others.

DiscussionStudy 1 provides initial evidence of the effects of specific

positive emotions on different types of prosocial behavior.Dispositional love (but not hope) predicts behaviors thatbenefit distant others, whereas both love and hope similarlypredict behaviors for close others. However, Study 1 hassome limitations. Specifically, we measured emotion ratherthan manipulating it, and we were unable to assess socialconnection directly. In the next three studies, we directlymanipulate emotion and measure social connection to pro-vide a more stringent test of our hypotheses.STUDY 2: LOVE AND HOPE LEAD TO DIFFERENT

PATTERNS OF PROSOCIAL CONSUMPTIONBEHAVIOR FOR CLOSE AND DISTANT OTHERSStudy 2 tests whether momentary experiences of love and

hope lead to different patterns of prosocial consumption.Specifically, we designed Study 2 to test whether incidentallove would increase intentions to perform prosocial behav-iors that benefit distant others more than incidental hope.Again, we expect prosocial behaviors that benefit close oth-ers to reveal general (i.e., undifferentiated) effects of posi-tive emotion on helping (i.e., both hope and love shouldlead to greater intentions to perform behaviors that benefitclose others than the neutral condition). However, love (butnot hope) should increase prosocial behaviors that benefitdistant others.Method and ProcedureEmotion induction pilot study. Sixty-five students com-

pleted a comprehensive pretest of the emotion inductionprocedure, self-reflective writing, which has been used suc-cessfully in many studies (e.g., Labroo and Rucker 2010;Lerner and Keltner 2001; Small and Verrochi 2009). Partici-pants were randomly assigned to one of four emotion condi-tions (hope, love, pride, or neutral) and asked to answer twoquestions on the computer. First, depending on the condi-tion, they were asked to describe three to five situations thatmade them feel a focal emotion (hope, love, or pride3) andto write two to three sentences about each situation. Next,participants were asked to describe in more detail the onesituation that made them feel the most of the focal emotionby typing a description of that situation. Those in the neutralcondition were asked to describe everyday activities in aformat designed to match the detail and length of the posi-tive emotion inductions (Lerner and Keltner 2001).Following the emotion induction, pilot study participants

completed multi-item manipulation check measures forarousal (stimulated and energized; a = .83); happiness(happy, joy, elated; a = .90); hope (hopeful, optimistic; a =.85); love (love, affection; a = .93); and pride (proud, confi-dent; a = .90) on a nine-point scale (0 = “none,” and 8 =“more than ever”) based on previously developed measuresfor assessing specific emotions (Dunn and Schweitzer 2005;Fredrickson et al. 2003; Rottenberg, Ray, and Gross 2007).The results revealed successful emotion induction withclean separation of the focal emotions (see Table 1). Com-

mon themes for the hope, love, and pride stories includedacademic and career goals, friends and family members,and competitive accomplishments, respectively (for writingsamples, see Table 1 and the Web Appendix). In the mainstudy described next, we manipulated emotion only at thelevels of love, hope, and a neutral emotional state. Weaddress pride in Studies 3 and 4.Participants and setup. For the main study, we used a 3

(emotion [between subjects]: love, hope, neutral) ¥ 2 (socialdistance of beneficiary [within subject]: close, distant)mixed design. Participants were randomly assigned to oneof the three emotion conditions that were pretested in thepilot study. Seventy-four university students participated ina study on feelings and consumer choice. The sample con-sisted of 45 men and 29 women ranging in age from 18 to30 years (M = 20.97 years, SD = 2.62). To dissociate theemotion procedure from the dependent measures of interest,participants were told that they would be completing a multi-part study that consisted of (1) a writing exercise on emo-tional experience, (2) a consumer choice survey, and (3)measures of their beliefs and opinions. Note that the emo-tion induction procedure was identical to the directed writ-ing procedure described in the pilot study.4Prosocial consumption behavior intentions. After the

emotion induction procedure, participants completed apaper-and-pencil consumer choice survey. This survey con-sisted of the same behavioral intention measures used inStudy 1.


3Pride was not part of the main Study 2 design. We included it in thepretest in the interest of manipulations for other studies.

4Because our conceptualization was based on companionate love, weexcluded participants who wrote about desire/passion (i.e., nudity and sex).Across studies, all participants in the love condition wrote stories aboutcompanionate love with the exception of five participants (n = 5) in Study 2and two participants (n = 2) in Study 3 who were identified by an indepen-dent coder and removed before analysis. The pattern of results is consistentif these participants are included; thus, we do not discuss them further.

Table 1EMOTION INDUCTION PILOT STUDY MEANS FOR EMOTION

MANIPULATION CHECKS

ValenceEmotion (“Happiness”) Arousal Hope Love PrideLove 6.3a 4.7a 6.0a 7.4b 6.9bHope 6.9a 5.4a,b 7.7b 5.1a 6.6bPride 6.3a 6.3b 6.5a 5.2a 7.8cNeutral 5.6b 5.2a,b 5.5a 4.5a 5.1aF 2.16 2.78 4.72 5.97 9.11p-value p < .10 p < .05 p < .005 p < .001 p < .0001Notes: Different superscript letters (a, b, c) within a given column indi-

cate significant differences at the level of at least p < .05. Emotion checkitems for valence (happiness), arousal, hope, love, and pride, were mea-sured on a nine-point scale. Example stories include the following: love: “Ifeel the most love when I receive a phone call out of the blue from an oldfriend I haven’t talked to in a while. It feels great because I know that myfriendship means something to them…”; hope: “I hope that I can travelwith my friends and enjoy this last opportunity before real life starts. I willhave to put a lot of effort into the medical school admissions process…”;pride: “I feel the most pride when thinking of my academic achievementsthroughout my whole education thus far. Academics have always beenimportant to me…”; neutral: “First I check my planner to see what home-work I have to do when I get home from class. I go to eat dinner at 5:45unless I have a meeting…” The Web Appendix provides additional partici-pant writing samples.


Social connection measure. After completing the depen-dent measures, participants answered questions about theemotion story written earlier in the study session. They indi-cated the extent to which a series of statements describedwhat they were feeling while writing (11-point scale; 1 =“not at all,” and 11 = “extremely”). We measured socialconnection using three items (a = .87; “To what extent did[the situation or experience you wrote about] affect the wayyou thought about your relationship with some individual orgroup?” “To what extent did you feel connected to anotherindividual or group?” and “To what extent did you feelclose or closer to another individual or group?”). We aver-aged these items to create a social connection score. Finally,participants completed demographic measures (age, gender,and ethnicity).5ResultsPreliminary analyses. Analysis of variance tests on the

social connection scores (F(2, 71) = 13.29, p < .0001)revealed significant emotion effects. Participants in the lovecondition experienced significantly more social connectionthan those in the hope (Mlove = 7.5, Mhope = 5.3; F(1, 71) =12.13, p < .0009) or neutral (Mneutral = 4.4, F(1, 71) = 25.64,p < .0001) conditions.Hypothesis tests. First, we examined H1 and H2, which

predicted an emotion by social distance interaction, withlove differentially increasing prosocial consumption behav-iors that benefit distant others relative to those that benefitclose others. In testing the likelihood of engaging in proso-cial consumption behaviors, we found a significant effectfor emotion (Mlove = 4.8, Mhope = 4.3, Mneutral = 4.2; F(2,71) = 3.96, p < .02) and a significant effect for social dis-tance (Mclose = 5.3, Mdistant = 3.5; F(1, 71) = 163.31, p <.0001), reflecting a higher likelihood of prosocial behaviorsfor closer beneficiaries. More interestingly, and as we pre-dicted, emotion significantly moderated the effect of socialdistance on likelihood to perform prosocial consumptionbehaviors (F(2, 71) = 5.32, p < .007). Both those in the love(F(1, 71) = 7.33, p < .01) and hope (F(1, 71) = 6.15, p < .02)conditions expressed significantly higher likelihoods ofprosocial consumption that benefits close others than those inthe neutral condition (Mlove = 5.54, Mhope = 5.48, Mneutral =4.86), in support of H1. Thus, with close others, we find anundifferentiated effect of positive emotions on prosocialbehaviors, consistent with Study 1 as well as prior research.Also as we predicted, however, those in the love conditionexpressed significantly higher likelihoods of prosocial con-sumption that benefits distant others than those in the hope(Mlove = 4.04, Mhope = 3.11; F(1, 71) = 8.53, p < .005) orneutral (Mneutral = 3.50; F(1, 71) = 3.86, p < .05) conditions,in support of H2; hope and neutral were equivalent (F(1, 71) =1.59, n.s.; see the Web Appendix). There was no effect ofemotion condition on intention to engage in filler itembehaviors (F(2, 71) = .24, n.s.; Mlove = 4.75, Mhope = 4.82,Mneutral = 4.64).

Next, we examined our social connection hypothesis (H3)that the impact of love on behaviors that benefit distant (butnot close) others is mediated by feelings of social connec-tion. Using the recommended technique for testing condi-tional indirect effects (Hayes 2013), process analyses(Model 14) confirmed evidence of moderated mediation.The effect of love on distant behaviors was mediated bysocial connection. We tested this using Hayes’ (2013)PROCESS macro with 5,000 bootstrapped samples. To testmediation of a three-group independent variable (Hayes2013, p. 196), we constructed two dummy variables, X1 andX2, representing the neutral and hope conditions, respec-tively. Because there were three groups, there are two indi-rect effects: (1) the indirect effect of a neutral emotion versuslove on distant behaviors through social connection and (2)the indirect effect of hope versus love on distant behaviorsthrough social connection.6 The indirect effect of neutral emo-tion versus love was B = –.4047 (SE = .1713), with a 95%bias-corrected bootstrapped confidence interval (CI) thatexcluded zero for distant behaviors (95% CI [–.7876, –.0906])but not close behaviors (B = .0039, SE = .1525, 95% CI = [–.3038, .2965]). The indirect effect of hope versus love wasB = –.2862 (SE = .1358), with a 95% bias-corrected boot-strapped confidence interval that excluded zero for distantbehaviors (95% CI = [–.6059, –.0653]) but not close behav-iors (B = .0027, SE = .1118, 95% CI = [–.2146, .2350]), insupport of H3. These findings provide evidence that themediational path predicting behavior from emotion is con-ditioned on the social distance of the beneficiary.Content analyses of emotion stories. We used content

analyses to test whether the difference found between emo-tion conditions could be attributed to differences in emo-tional intensity, cognitive processing style, or semanticpriming. Participants’ written passages were analyzed withtextual analysis (Linguistic Inquiry and Word Count [LIWC;http://www.liwc.net/]). These analyses showed statisticallyinsignificant effects of emotion condition for magnitude-of-emotion words and words related to cognitive processes inthe stories, suggesting that differences between conditionsare not driven by emotional intensity or cognitive process-ing style. Love is associated with more social words, but thefrequency of social words does not mediate our behavioraleffects, casting doubt on semantic priming of social pro-cesses as an alternative explanation (for statistical analysesand results, see the Web Appendix; for further elaboration,see the “General Discussion” section).DiscussionStudy 2 demonstrates that the specific positive emotions

of love and hope influence prosocial consumption thatbenefits close versus distant others in different ways. Wepredicted that love, characterized by social connection,increases intentions of engaging in behaviors that benefitdistant others more than hope, which is lower in social con-nection; our results support our prediction. In addition,social connection mediates the relationship between emo-tion and behaviors that benefit distant others. In contrast,5Initial analyses revealed a significant main effect for ethnicity, which

did not interact with the manipulations. Specifically, ethnic minorities indi-cated a greater propensity to perform prosocial consumption behaviorsregardless of emotion condition. However, although the results reportedherein do not include an ethnicity covariate, including such a covariatedoes not affect or qualify the results.

6As Hayes (2013) outlines, running PROCESS twice—once with X1 asthe IV and X2 as the covariate and once with X2 as the IV and X1 as thecovariate—enables the researcher to recover each indirect effect.

both positive emotions (love and hope) increase intentionsto perform behaviors that benefit close others over a neutralemotional state, replicating established findings that posi-tive emotion generally increases prosocial behavior forclose others.Given the particular emotions contrasted in Studies 1 and

2, one might argue that hope may be characterized by aunique quality that could explain the difference in reportedbehaviors. To address this concern, we compare love with adifferent specific positive emotion (pride) in Study 3. Inaddition, when comparing emotions it can be difficult toequate strength, and perhaps love tends to be experiencedmore strongly or as being more positive. To directly addressthe question of whether magnitude of positivity could pre-dict our findings, we measure and control for positivity inStudy 3, allowing for a more stringent test of differentialbroadening effects. We also employ a different prosocialcontext intimately linked to marketing (i.e., fundraising) forgreater generalizability and examine decisions with realdonation consequences.

STUDY 3: LOVE (NOT PRIDE) INCREASESDONATIONS TO DISTANT OTHERS

Study 2 demonstrates that two specific positive emotions(love and hope) differentially influence engagement inbehaviors that benefit distant others. Study 3 tests a differ-ent pair of specific positive emotions (love and pride) usinga fundraising context. We replicate and extend our socialconnection findings by showing that love and pride differ-entially influence to whom (e.g., domestic vs. internationalfunds) consumers give. Note that in Study 2 participantscould choose as many prosocial behaviors as they wanted(i.e., no explicit trade-off was required). We designed theStudy 3 task so that participants had to decide between bene-ficiaries (i.e., whether to give the most help to close or distantothers), providing a more rigorous test of our hypothesis.Nonprofit appeals often describe warm moments shared

between aid recipients and organizers or depict the proudfaces of volunteers who have worked to build homes andclinics, leading consumers to experience different specificemotions (love vs. pride). Could these distinct emotionalexperiences cause consumers to give in different ways? InStudy 3, we conceptually replicate our Study 2 findings,which suggest that love, characterized by social connection,will increase the likelihood of giving to international relieffunds, whereas pride, an emotion not characterized bysocial connection, will not. In Study 3, we focus on mone-tary giving, holding both charitable organization and overallamount given constant, to better understand consumers’ pri-oritization of beneficiaries. Participants responded to thefundraising appeal with the understanding that their dona-tion decisions had real behavioral consequences (i.e.,money would actually be given to the American Red Crossin the way that they designated).Method and ProcedureParticipants and setup. One hundred seventy-six students

completed a study on feelings and consumer choice. Thesample consisted of 111 men and 65 women ranging in agefrom 18 to 29 years (M = 20.50 years, SD = 1.12). To disso-ciate the emotion procedure from the dependent measures

of interest, participants were told that they would be com-pleting a series of short studies from different researchersthat had been bundled. The study session consisted of threeparts: (1) a writing exercise on autobiographical experience,(2) responding to a fundraising appeal, and (3) a question-naire about feelings and consumption behaviors.In Study 3, participants were randomly assigned to one of

the three emotion conditions (emotion: love, pride, neutral).After completing the same emotion induction procedure asin Study 2 for the focal emotions in the current study (i.e.,love, pride, and a neutral emotion), all participants viewedthe same fundraising appeal from the American Red Cross.Adapted from actual materials, the fundraising appealdescribed the organization’s activities and ways in whichdonors could give to the organization. Participants wereasked to make their donation decisions as they really wouldat this moment and were explicitly told that we would select“1 out of every 20 participants’ decisions and actuallydonate to the American Red Cross” as they had specified.Domestic versus international relief fund donation deci-

sion. The donation form that participants received listed twooptions: a domestic and an international relief fund, bothdescribed as providing immediate relief from suffering andlong-term support. The order in which these two funds werelisted was counterbalanced across participants; we found noorder effects. Participants were asked, “If right now you had$50 to donate, how would you allocate your donation?”They then entered an amount ($0–$50) in each of the spacesprovided. The dependent measure was total dollars allo-cated to international relief.Social connection and emotion check. After completing

the dependent measures, social connection was measured asin Study 2. Participants were also asked to revisit their sto-ries and recall specifically how they were feeling whenwriting them. Participants rated the extent to which they feltthree positive emotions (i.e., happiness, love, and pride) on11-point scales (1 = “not at all,” and 11 = “extremely”) aswell as the magnitude of these emotions (e.g., “How muchlove did you feel?” “How much pride did you feel?”) on a7-point scale (1 = “not at all,” and 7 = “very much”). Wecalculated a measure of positivity (i.e., the average of allthree positive emotion items) for each participant to controlfor general positivity (see Oveis, Horberg, and Keltner2010). Finally, participants completed demographic mea-sures: age, gender,7 and ethnicity.8ResultsPreliminary analyses. As recommended in Oveis, Hor-

berg, and Keltner (2010), we included a general positivitymeasure to control for magnitude of positivity as a possiblealternative explanation. Initial analyses revealed a signifi-


7Recent findings have suggested that men and women may respond dif-ferently to donation requests involving in-groups and out-groups (Win-terich, Mittal, and Ross 2009); however, we found no gender differences inlikelihood of giving to domestic versus international relief funds in our stud-ies. Moreover, gender did not moderate the effect of emotion on donations.8Consistent with Study 1, ethnic minorities reported a greater propensity

to give to distant others (i.e., international funds). Again, the pattern ofresults and significance of comparisons remain the same with or withoutethnicity included. The results we report herein do not contain an ethnicitycovariate.


cant main effect for positivity. This effect, however, did notmoderate our results. Participants reporting higher generalpositivity indicated a greater propensity to give to close oth-ers (i.e., domestic funds).Drawing on participants’ responses to the social connec-

tion items (a = .88), we again created a social connectionscore for each participant. Tests on social connection scores(F(2, 172) = 12.53, p < .0001) revealed significant emotion-specific effects. Participants in the love condition reportedgreater social connection than those in the pride (Mlove =8.34, Mpride = 6.54; F(1, 172) = 19.49, p < .0001) or neutral(Mneutral = 6.32; F(1, 172) = 18.60, p < .0001) conditions.Hypothesis tests. Examination of dollars donated to inter-

national versus domestic relief funds enabled us to test ourhypothesis, which predicts that love will lead people to givemore to international relief. In a model predicting dollarsdonated to international relief, we found a marginally sig-nificant effect for emotion (F(2, 172) = 2.66, p < .07). Peo-ple experiencing love donated significantly more money tointernational relief than those experiencing pride (Mlove =$20.22, Mpride = $14.56; F(1, 172) = 3.97, p < .05) or thosein a neutral emotional state (Mneutral = $13.60; F(1, 172) =4.11, p < .04).We tested for mediation by social connection using

PROCESS analyses (Hayes 2013)9 and requested estimatesof the conditional indirect effects at different levels of posi-tivity. The indirect effect of neutral versus love, with 95%bias-corrected bootstrapped confidence interval, was sig-nificant at all levels of positivity (–1 SD: B = 1.3447, SE =.8623, 95% CI = [.1310, 3.7795]; mean: B = 1.8802, SE =1.0431, 95% CI = [.0834, 4.2751]; +1 SD: B = 2.4157, SE =1.3933, 95% CI = [.1094, 5.6659]). The indirect effect ofpride versus love was also significant across all levels ofpositivity (–1 SD: B = 2.4386, SE = 1.4328, 95% CI =[.1119, 5.8799]; mean: B = 1.8290, SE = 1.0411, 95% CI =[.0379, 4.1958]; +1 SD: B = 1.2193, SE = .8057, 95% CI =[.0725, 3.4443]). These results provide evidence that love’seffect on monetary donations to international relief is medi-ated by social connection, regardless of the magnitude ofpositivity experienced. Moreover, they provide further evi-dence that differences found between emotion conditionsare attributable to differences in social connection and notemotional intensity.10Content analyses of emotion stories. We conducted the

same LIWC analyses described in Study 2. Our statisticalanalyses showed no evidence for the alternative mecha-nisms discussed. The effects were either insignificant (emo-

tional intensity) or opposite in direction (cognitive process-ing style). Again, love was associated with more socialwords, but the frequency of social words did not mediateour behavioral effects, casting doubt on semantic priming ofsocial processes as an alternative explanation (for detaileddescriptions and results of our statistical analyses, see theWeb Appendix).DiscussionStudy 3 demonstrates that specific positive emotions lead

to giving to different types of recipients (i.e., domestic vs.international relief funds). People feeling love are morelikely to give money to international relief than people feel-ing pride. Again, Study 3 demonstrates that love affectsbehavior in a way that hope and pride do not. In an addi-tional study (Study 3b), we replicated this result for loveusing a different comparison emotion (hope) but the sameprocedure, with the exception of adding multi-item emotionchecks. Participants in the love condition reported signifi-cantly more love, less hope, and more social connectionthan those in the hope condition (for details, see the WebAppendix). In a model predicting dollars donated to inter-national relief, we found a significant effect for emotion(F(1, 36) = 7.98, p < .008). People experiencing lovedonated significantly more money to international reliefthan those experiencing hope (Mlove = $19.89, Mhope =$11.98).Thus, the emotional state of potential donors does not

simply influence whether they give but, more specifically,to whom they give—close versus distant others (i.e., domes-tic vs. international funds). These findings have importantimplications for universities and nonprofit organizations,which regularly allow donors to decide how to direct mone-tary gifts, and they also underscore the importance of char-acterizing positive emotions on the basis of their uniquebroadening properties.Studies 2 and 3 show that social connection matters;

however, is social connection alone sufficient for broaden-ing? Might the negative feelings that accompany compas-sion dampen giving to distant others? We hypothesize thatthis will be the case—that is, that the co-occurring positiveand negative feelings characteristic of compassion will notlead to the broadening tendency fostered by love with itscharacteristic positive feelings. Study 4 addresses this pre-diction by including compassion along with all the emotionconditions used previously. We also test our hypothesesusing a different study approach that is more relevant tomarketing communications. Studies 2 and 3 use an estab-lished procedure (writing about a personal experience) toinduce emotion and measures of behavioral intentionstoward close and distant others (Study 2) and monetary giv-ing toward domestic versus international relief funds (Study3) within one organization (American Red Cross). Toincrease confidence in the validity and generalizability ofour findings, Study 4 provides additional evidence using amore naturalistic emotion manipulation and different chari-table organizations. We use magazine advertisements tomanipulate emotions and then ask participants to make amonetary giving decision involving two different charitableorganizations while allowing them a realistic third option ofkeeping money for themselves.

9We conducted mediation analyses using PROCESS (Hayes 2013),which allows model estimation using three or more conditions; Model 7was specified with 5,000 bootstraps.10We also tested whether the causal structure we identify can be distin-

guished from one in which love and social connection are switched to pre-dict giving to distant others. In the alternative model, the coefficients forsocial connection (B = .22, t = .41, p = .68) and reported love (B = .59, t =.46, p = .65) were not significant. Moreover, the indirect effect of socialconnection was B = .10 (SE = .22), with a 95% bias-corrected bootstrappedconfidence interval that included zero (95% CI = [–.3355, .5672]). In sum-mary, we find that although higher levels of social connection are associ-ated with feeling more love (B = .18, t = 5.93, p < .001), social connectiondoes not directly predict behavior without antecedent love, and the indirecteffect of social connection on behavior through love does not yield supportfor the alternative causal structure.

STUDY 4: LOVE DIFFERENTIALLY PREDICTS GIVINGTO INTERNATIONAL (VS. LOCAL) CHARITIES

We designed Study 4 to provide further evidence forlove’s effect on behaviors that benefit distant others. We useadvertisements for one brand (Canon) to manipulate fiveincidental emotions (love, hope, pride, compassion, and aneutral emotion). Including compassion enables us to con-currently examine three social emotions (love, pride, andcompassion) in addition to hope and to assess whether thecaretaking motives associated with compassion may predictthe same effects as love or whether love’s broadeningeffects are unique. Although Study 3’s donation allocationhad real consequences (i.e., money given to the AmericanRed Cross), it required participants to choose between bene-ficiaries (i.e., they did not have the option to keep themoney). To address this limitation and further increase real-ism, we included a third option (i.e., keep money for one-self) along with the two different charity organizationoptions: after completing the advertisement evaluation, in aseparate task, participants were provided with an opportu-nity to donate the money toward two different environmen-tal organizations (local vs. international) or to keep themoney. We again focus on relative aid to distant (inter-national) versus closer (local) others, but in this case we cal-culate it in the context of money that people freely chose togive. Specifically, we designed Study 4 so that participantscould decide exactly how much they wanted to give to alocal environmental group, give to an international group,or keep for themselves (i.e., we made the trade-off appar-ent). This design provides an additional test of our predic-tion that love causes a shift toward behaviors that benefitdistant others.Emotion Induction Pilot Study for AdvertisementsWe developed our stimuli with extensive pretesting and a

comprehensive pilot study to obtain valid emotion checks(Herr et al. 2012, p. 835). One hundred forty-seven studentsfrom the same population as those in the focal study partici-pated in a comprehensive pretest of the emotion inductionprocedure. Participants were told that they would be com-pleting a study about magazine advertising and that we wereinterested in consumers’ reactions to and memories foradvertising. They were randomly assigned to one of fiveemotion conditions: compassion, hope, love, pride, or neu-tral. Each participant viewed one Canon camera ad. Acrossconditions, the layout and slogan (“Capturing MomentsLike This”) were held constant, but the image and bodycopy varied with emotion condition (see Table 2 and theWeb Appendix). We selected all images for the ads in linewith extensive pilot testing with online samples.11 Weadapted the body copy for each emotion from prototypicalthemes and moments shared in emotion stories written byparticipants in our previous studies. Notably, the emotion-manipulating advertisements did not contain any referenceto the focal emotions being manipulated. This absence ofthe actual emotion terms (i.e., love, hope, pride, and com-passion) provides a conservative, cleaner test compared

with what is likely done in practice, whereby advertisers canand do use the actual emotion words as well.After viewing the ad, pilot study participants indicated

the extent to which they experienced a series of specificfeelings while viewing the advertisement on a seven-pointscale (1 = “did not experience at all,” and 7 = “experiencedvery intensely”). Each of the focal emotions was assessedwith three items using terms reported previously in the emo-tions literature: compassion (“compassion,” “sympathy,”“moved”; a = .80), hope (“hopeful,” “optimistic,” “encour-aged”; a = .87), love (“love,” “affection,” “closeness”; a =.86), pride (“proud,” “achievement,” “self-assured”; a = .81),and neutral (“neutral,” “unemotional,” “indifferent”; a =.86). We ran a model with the emotion condition and thepositivity score entered as predictor variables to assesswhether the magazine ads effectively manipulated specificemotions (for details, see Table 3). In summary, the resultsshow that our ad manipulation cleanly differentiated lovefrom all other emotions. In addition, each emotion adelicited significantly more of the intended focal emotionthan any other emotion ad, with one exception (pride).12Participants were shown the same ad a second time and thenrated their ad-related feelings using single-item measuresfor the extent to which they had a positive or negative emo-tional response and how emotional they felt while viewingthe advertisement on seven-point scales (1 = “not at all,”and 7 = “very”). We found no significant differencesbetween the love ad and the other emotion ads for thesingle-item measures, with one notable exception: the com-passion ad. Planned pairwise comparisons with the love ad


11Oveis, Horberg, and Keltner (2010) previously validated the compas-sion image.

12Those in the pride ad condition reported significantly more pride thanany other condition; however, they also reported substantial hope. Providedthat people view the accomplishments in the pride condition as desirable, itis not entirely surprising that reading about success would also instill hopefor that type of success or hope for success more generally. We note that wepretested numerous iterations of a pride advertisement, varying the imageand statements used in the body copy. Within an advertising paradigm, thecoactivation of hope with pride was recurrent despite numerous attempts tofully isolate pride. The isolation of pride from hope is more readily appar-ent within a writing paradigm, in which the experiences recalled are indi-viduating (i.e., “I can recall an experience that has made me feel the mostpride”), whereas the advertisement is meant to induce pride by drawing onprototypical but hypothetical situations that may or may not be accessibleor applicable to that person.

Table 3MAGAZINE ADVERTISEMENT PILOT STUDY MEANS FOR

EMOTION MANIPULATION CHECKS

Emotion AdCondition Love Hope Pride Compassion NeutralLove 4.6d 3.3a 2.2a 3.6b 3.1aHope 3.4b 4.3c 3.1b 3.8b 3.2aPride 2.8a 4.2c 4.0c 2.9a 3.3aCompassion 4.2c 3.5a,b 2.5a 4.6c 3.0aNeutral 3.4b 3.8b 3.5b 3.3a,b 4.7bF 32.22 8.69 18.39 17.96 4.99p-value p < .0001 p < .0001 p < .0001 p < .0001 p < .001Notes: Different superscript letters (a, b, c, and d) within a given column

indicate significant differences at the level of at least p < .05. Boldfaceindicates that each of the emotion-specific ads elicited the highest level ofthe focal emotion. Emotion check items were measured on a seven-pointscale.


Love

Pride

Hope

Comp

assion

Neutral

Advertisement

Body Copy

Every tim

e a baby girl holds fast to her

father’s thumb. Every tim

e a young boy

picks flowers for his mom. Every time a

mother bakes chocolate chip cookies for

her daughter. Every time a gray-haired

couple sits elbow-to-elbow at a picnic

table sharing sm

iles and watermelon

slices. Every tim

e a furry best friend

greets her person at the door. E

very time

you visit your oldest and dearest friends

and reunite in a warm

embrace.

Every tim

e an athlete claim

s first place.

Every tim

e an actor gives a

performance that brings the house

down. Every time a grandmother bakes

her secret-recipe apple pie for a contest

and it wins the grand prize. Every tim

ea scholar completes her graduate

education having achieved the school’s

highest honors. Every time a runner

raises his arms in victory because he

crossed the finish line with a personal

best. Every time you have beaten out

all the other applicants for a prized job.

Every tim

e a little girl dreams she could

earn a spot in the ballet company. Every

time a couple puts a picture of a house

on their apartment refrigerator thinking

it may someday be in reach. Every time

a little boy sets up a lem

onade stand

and tries to earn enough money for that

awesome bicycle. Every time a student

writes a personal essay for her

application and then waits and wonders

about acceptance to the university her

heart is set on. Every tim

e an actor,

knowing it is a long-shot, prepares to

audition for the role of his lifetime.

Every tim

e you have a chance to

interview

for your dream job, knowing

the competition is steep.

Every tim

e a soldier comforts his

wounded brother-in-arms. Every tim

e avolunteer in a nursing home reads the

newspaper to an elderly person w

hocannot see. Every time someone sees a

homeless person sitting in the rain and

stops and gives away her umbrella.

Every tim

e friends come together to

assist with care-giving during the last

months of their neighbor’s life. Every

time a student notices a new kid in the

cafeteria and invites him

to sit at his

table for lunch. Every tim

e you take the

time to help someone w

ho is in need.

Every tim

e a student walks across

campus on the w

ay to her regular

weekly classes. Every tim

e a comm

uter

reads the newspaper on a daily bus ride

to and from

work. Every time a

businessman crosses through a street

intersection on the way to an office

downtow

n. Every tim

e a staff person sits

in front of the computer working on a

document at her desk. Every time a man

shops for his weekly groceries at the

supermarket. Every time you walk down

the sidewalk on your typical route.

Tagline

A moment like this… reminds you of

dear ones and touches your heart.


having achieved your very best.

A moment like this… reminds you that

reaching your goals may be possible.


kind acts toward those most in need.


the normal situations that make up your

days.

Table 2

ADVERTISEMENTS USED FOR INDUCTION OF SPECIFIC POSITIVE EMOTIONS

revealed that the compassion ad was viewed as less positive(Mlove = 4.81, Mcompassion = 3.93; F(1, 137) = 11.34, p <.001), more negative (Mlove = 2.07, Mcompassion = 3.29; F(1,137) = 11.98, p < .001), and more emotional (Mlove = 3.26,Mcompassion = 4.21, F(1, 137) = 8.48, p < .004) than the lovead.Main StudyParticipants and setup. For the main study, two hundred

six students completed a multipart advertising study forcourse credit. The first task in the study was affectively neu-tral and constant across conditions. Its purpose was to neu-tralize affect and reinforce the cover story. First, all partici-pants completed the same “image and graphics pretest” inwhich they viewed and rated three neutral images (pens, sta-pler, and outdoor space; White, Kenrick, and Neuberg 2014)in terms of how they would rate an ad that used this picture.The second task (“magazine advertising”) provided themanipulation of specific emotions. The dependent measurewas collected in a third consumer decision-making task. InStudy 4, participants were randomly assigned to one of thefive emotion conditions (emotion: love, hope, pride, com-passion, neutral), which were pretested in the aforemen-tioned pilot study.Emotion induction. Each participant viewed one of the

five ads described in the pilot study. Importantly, unlike thepilot study they were not asked about their feelings becauselabeling one’s feelings after incidental emotion inductionscan reduce the effect of such emotions (Cryder et al. 2008;Schwarz and Clore 1983).Domestic versus international environmental fund dona-

tion decision. After completing the magazine advertisingtask, participants completed a consumer decision task, inwhich they were asked to make a donation decision “as youreally would at this moment.” They were told to imaginereceiving a $10 bonus payment and that they could eitherdonate to the following nonprofit organizations helping theenvironment or keep the $10 bonus payment. Participantswere asked how they chose to allocate the bonus paymentand entered amounts ($0–$10) for each of the followingoptions: “Environmental Defense Local Fund,” “NaturalResource Federation International Fund,” and “Keep forSelf.” We conducted a charities pretest (N = 66) before themain study to ensure that both charities were equally pre-ferred. Paired comparisons indicated that participants gaveequivalent amounts to both the local (M = $2.58) and inter-national (M = $2.50) environmental charities (t(65) = .17,n.s.) but kept significantly more for themselves (M = $4.92)than they gave to either the local (t(65) = –2.66, p < .01) orthe international (t(65) = –2.79, p < .007) groups absent anyemotion manipulation.Our focal prediction (H2) is that love will increase the

propensity to give to the international fund, shifting thefocus from general tendencies to prefer local giving. To testwhether love shifts priority toward more socially distantbeneficiaries, we used these allocation amounts to createboth a giving index (international – local) and a proportiongiven to international relief (international/total amountdonated) for each participant, which served as the depen-dent measures. Note that we can only calculate the propor-

tion measure for those who chose to donate something (67participants chose to keep all the money); the giving indexenables us to use data for all the participants, ensuring thatour proportion results are not somehow an artifact of shiftsin propensity to opt out of giving at all.13Social connection, emotional response, and background

measures. Participants completed the same social connec-tion measures (a = .93) as in Studies 2 and 3. They alsocompleted a short series of questions about the advertise-ments they had viewed, indicating the extent to which theywould rate their emotional response as positive and as nega-tive and the extent to which they felt emotional when view-ing the advertisement on seven-point scales (1 = “not at all,”and 7 = “very”). Finally, participants completed basicdemographic measures (age, gender, and ethnicity).Hypothesis tests. To determine whether the expected dif-

ference emerged in terms of whether the funds were allo-cated to the international versus local fund (i.e., distant vs.close), we ran analyses of variance predicting the proportiongiven to international relief from emotion condition and thegiving index. Consistent with our theorizing, when donationsare analyzed as a proportion (international/total amountdonated), the overall model is significant (F(4, 133) = 2.45,p < .05), and those who viewed the love ad (M = .58) allo-cated a larger proportion toward the international group thanthose in the compassion (M = .36; F(1, 133) = 8.67, p < .004),hope (M = .43; F(1, 133) = 4.29, p < .04), pride (M = .45;F(1, 133) = 3.62, p < .06), or neutral (M = .42; F(1, 133) =4.87, p < .03) ad groups. In addition, we find a significanteffect of emotion on the giving index (F(4, 200) = 2.57, p <.04). Those who viewed the love ad (M = .79) allocatedrelatively more toward the international group than those inthe compassion (M = –1.15; F(1, 200) = 8.72, p < .004),hope (M = –.64; F(1, 200) = 4.54, p < .03), pride (M = –.15;F(1, 200) = 2.08, p < .15), or neutral (M = –.75; F(1, 200) =5.34, p < .02) ad groups14 (see the Web Appendix).We also conducted follow-up analyses on the supplemen-

tal measures. As we expected, participants reported experi-encing greater social connection with the love ad (Mlove =8.45) than the hope (Mhope = 7.19; F(1, 200) = 8.11, p <.005), pride (Mpride = 6.44; F(1, 200) = 21.94, p < .0001),and neutral (Mneutral = 5.93; F(1, 200) = 32.05, p < .0001)


13We calculated the proportions of those donating at all, donating to thelocal charity, and donating to the international charity; we found no signifi-cant differences across conditions (p = .24, p = .43, and p = .18, respec-tively). We also calculated how much money participants donated in total.We found no differences in this total (F < 1) across conditions; thus, we donot discuss total donations (and the linearly related amount kept for the self[$10 – amount donated]) further.14As a robustness check, we also examined amounts allocated to the

international and local funds using seemingly unrelated regression, runningthe SYSLIN procedure in SAS. The two equations were estimated simulta-neously, along with the cross-model correlation, using four conditiondummy variables (for five emotion conditions). We found that love (B =1.25, t(200) = 2.31, p < .02) was a significant predictor of amount donatedto the international fund, but compassion (B = –.66, t(200) = –1.26, p <.21), hope (t < 1), pride (t < 1), and a neutral emotional state (t < 1) werenot. In contrast, compassion (B = 1.28, t(200) = 2.33, p < .02), hope (B =1.05, t(200) = 1.88, p < .06), pride (B = .88, t(200) = 1.63, p < .10), and aneutral emotion (B = 1.55, t(200) = 2.78, p < .006) were predictors ofamount donated to the local fund, but love was not (t < 1). The cross-modelcorrelation for the seemingly unrelated regression analysis was .249, with a95% CI that excluded zero (.1114, .3866).


ads but reported equivalent social connection with the com-passion ad (Mcompassion = 8.23; F < 1). The love and com-passion ads also elicited emotional responses that wererated as similarly emotional (Mlove = 3.80, Mcompassion =3.63; F < 1) and positive (Mlove = 4.98, Mcompassion = 4.62;F(1, 200) = 2.53, p < .11). However, the compassion ad alsoelicited a greater negative emotional response than love(Mlove = 2.04, Mcompassion = 2.81; F(1, 200) = 7.37, p <.007). Notably, love’s effect is not explained by magnitudeof positive emotion alone (i.e., love, hope, and pride wererated as equally positive [all Fs < 1]).Holding compassion aside, we tested whether social con-

nection mediated the demonstrated effect. To test mediationof a four-group independent variable (Hayes 2013, p. 196),we constructed three dummy variables, X1, X2, and X3,representing the neutral, hope, and pride conditions, respec-tively. With four groups, there are three indirect effects: (1)the indirect effect of neutral emotion versus love on the giv-ing index through social connection, (2) the indirect effectof hope versus love on the giving index through social con-nection, and (3) the indirect effect of pride versus love onthe giving index through social connection.15 We tested formediation by social connection using PROCESS analysesand requested estimates of the conditional indirect effects atdifferent levels of positivity. PROCESS analyses (Model 14with 5,000 bootstrapped samples) confirmed evidence ofmoderated mediation. The effect of love on the giving indexwas mediated by social connection at moderate and high(but not low) levels of positivity. The indirect effect of neu-tral versus love, with a 90% bias-corrected bootstrappedconfidence interval, was significant at moderate and highlevels of positivity (–1 SD: B = .0043, SE = .3407, CI = [–.5807, .5258]; mean: B = .5881, SE = .4093, CI = [.0548,1.3989]; +1 SD: B = 1.1719, SE = .7442, CI = [.1695,2.6603]). The indirect effect of hope versus love was alsosignificant at moderate and high (but not low) levels of posi -tivity (–1 SD: B = .0015, SE = .1288, CI = [–.2390, .1834];mean: B = .2032, SE = .1568, CI = [.0247, .5797]; +1 SD: B = .4049, SE = .3060, CI = [.0521, 1.1502]). The indirecteffect of pride versus love was also significant at moderateand high levels of positivity (–1 SD: B = .0024, SE = .1907,CI = [–.3256, .3018]; mean: B = .3233, SE = .2303, CI =[.0302, .7975]; +1 SD: B = .6443, SE = .4154, CI = [.1130,1.5247]). These results provide evidence that love’s effecton the giving index is mediated by social connection atmoderate to high (but not low) levels of positivity. Note thatnot all these comparisons hold with 95% CIs.DiscussionStudy 4 shows that love (but not hope, pride, compassion,

or a neutral emotion) increases priority placed on giving tointernational charity organizations. Study 4 provides furtherevidence that love is distinct from hope and pride in termsof social connection. Although both love and compassionare associated with feelings of social connection, only lovecaused consumers to donate more to distant others. This evi-

dence suggests that love has a unique broadening effect andthat broadening through social connection is expected inresponse to socially connecting positive emotional experi-ences and not to those accompanied by the co-occurrence ofnegative feelings.

GENERAL DISCUSSIONWe examine whether and how specific positive emotions

can influence to whom consumers give resources. Ourresults show that, to date, the relationship between positiveemotion and prosocial behavior has been overgeneralized.Four studies show that specific positive emotions predictunique patterns of prosocial behavior. Studies 1 and 2 showthat love and hope influence prosocial consumption thatbenefits close and distant others in different ways. Disposi-tional (Study 1) and incidental (Study 2) love but not hopeincreases willingness to perform behaviors that benefit dis-tant others (e.g., refugee families). Whether measured ormanipulated, love increases behaviors that benefit distantothers, but both hope and love similarly affect behaviorsthat benefit close others. Study 3 shows that specific posi-tive emotions change how people give in response tofundraising appeals—specifically, the amount of moneygiven to different types of beneficiaries. Love increasesdonations to distant others (i.e., international relief funds)relative to pride, neutral emotion, and hope (Study 3 andreplication study). In addition, love increases donation allo-cations to international organizations relative to hope, pride,compassion, and a neutral emotion (Study 4). This relation-ship between specific positive emotions and behaviors thatbenefit distant others is explained in part by feelings ofsocial connection (Studies 2 and 3) but is also qualified bythe nature of the emotional experience (Study 4). Morespecifically, broadening through social connection to helpdistant others seems to require social connection predicatedon positive feelings (vs. co-occurrence of positive and nega-tive feelings).Theoretical ContributionsThis research has important implications for emotion

theory and offers the first empirical demonstration of differ-ential broadening, a unique contribution to the broaden-and-build theory of positive emotion (Fredrickson 1998, 2001;Fredrickson et al. 2008). Previous tests of the broaden-and-build theory have shown that all positive emotions broadenin similar ways by broadening attention, cognitive and moti-vational scope, and range of thoughts and actions. We findthat love, a positive emotion characterized by social connec-tion, leads to prosocial consumption behaviors that benefitdistant others as well as donations to international relieffunds and organizations, whereas other specific positiveemotions do not. This distinction that some positive emo-tions broaden in a way that others do not is a first in the lit-erature. Why have these effects not been found previously?To date, Fredrickson and colleagues (Frederickson andBranigan 2005; Frederickson et al. 2008; Frederickson et al.2003) have focused on demonstrating that positive emotionsas a group have functionality (i.e., broadening) distinct fromthe narrowing action tendencies promoted by negative emo-tions research. They have also focused on showing that

15As outlined by Hayes (2013), running PROCESS three times—oncewith X1 as the IV and X2 and X3 as covariates, once with X2 as the IV andX1 and X3 as the covariates, and once with X3 as the IV and X2 and X3 asthe covariates—enables the researcher to recover each indirect effect.

broadening effects are not a function of arousal. Thus, theirresearch does not address the array of positive emotions andmore detailed view of broadening we suggest.Our findings not only provide the first demonstration of

differential broadening but also contribute by highlightingan important dimension on which positive emotions differ.Previous research has focused largely on negative emotionsand dimensions important to their differentiation (i.e., cer-tainty and control; e.g., Lerner and Keltner 2001). Ourresearch suggests that positive emotions differ along animportant dimension related to broadening behaviors andprovides a richer understanding of the dimensions central todifferentiating positive emotions.Implications for Prosocial Consumption and MarketingOur research also has important implications for proso-

cial consumption and prosocial behavior more generally byhighlighting beneficiary focus as an important dimension ofprosocial consumption. Prosocial consumption behaviorsdiffer substantially in terms of the type of person or causebenefited. Our research demonstrates the implications ofspecific positive emotions for different types of beneficiar-ies. These findings suggest that marketers and policy mak-ers need to be keenly aware of the type of emotion they areusing in their appeals to consumers. In addition, our find-ings help illuminate why different levels of prosocial con-sumption behaviors may occur in different contexts (e.g.,when appeals contain different types of positive emotionalcontent).It is not the case that one positive emotion (love) is uni-

versally better at motivating all behaviors than others.Rather, the relative effectiveness of love in marketingdepends on the type of behavior desired. In a fundraising orsocial marketing context, a campaign that benefits distantothers will be more effective using love rather than hope,pride, or compassion. For close others, the set of positiveemotions may be equally effective. Managers and policymakers also may strategically try to increase potentialdonors’ and consumers’ sense of social connection to otherswithin the context of a nonprofit, a university, or a brand,particularly when the beneficiary or product is less familiarto them. Our findings underscore the importance of differen-tiating among a fuller spectrum of specific positive emotionswhen designing and testing persuasive communications.LimitationsAlthough our article offers evidence of love’s impact on

prosocial behaviors through feelings of social connection,we do not definitively resolve the cognitive and emotionalprocess that could possibly be producing the results. Forexample, empathy and other unmeasured constructs may beinvolved. In addition, our focus on love in companionaterelationships may imply that connectedness is a necessaryprecursor to love (for results testing the reverse mediation,see Study 3). Emotion researchers have long grappled withthe classic “chicken or egg” question regarding the relation-ship between cognition and emotion, with some theoristscontending that affect precedes cognition (e.g., Zajonc1980) and others contending that cognition precedes affect(e.g., Bower 1981). Recent work has suggested that affect

and cognition are highly interdependent (Storbeck andClore 2008). Further research is needed to shed light onsuch issues.In addition, our attempts to measure and manipulate emo-

tions relied on self-report data, which is a limitation of thisresearch. To manipulate specific emotions, we employed awritten emotion induction procedure requiring participantsto write about autobiographical events (Studies 2 and 3).This procedure is the most common emotion manipulationprocedure in the marketing literature (we surveyed emotionstudies appearing in Journal of Marketing Research andJournal of Consumer Research between 2003 and 2013 andfound that 60.4% used this method). Because this methodrelies on written statements, some may question whether itproduced felt emotion or simply activated or semanticallyprimed emotion-related words. We believe that our data donot support an account that results solely from priming.First, the detailed LIWC analyses of participants’ stories donot support a mere semantic priming account. Specifically,the frequency of social process–related words in the storiesdid not mediate the relationship between love and behaviorsbenefiting distant others (Study 2 and 3), and the other wordtypes analyzed either did not differ across the different posi-tive emotions examined or did not mediate the results. Sec-ond, the dispositional emotion results in Study 1 are diffi-cult to explain as being due to simple priming based onpresence of an emotion word (or words) in the writing taskinstructions and responses, given that in the fully within-subject design all participants read all of the same emotionwords. Third, the advertisement results in Study 4 aredemonstrated without the use of the actual emotion wordsever appearing in the advertisements. Finally, our detailedmanipulation check data suggest that participants reportexperiencing these emotions.Further ResearchIn this article, we endeavored to shed new light on a more

diverse set of specific positive emotions often used by mar-keters. Our efforts offer multiple avenues for furtherresearch, as we discuss next.Social connection. Our approach offers a framework for

making additional predictions linking sp

lisa a. cavanaugh, james r. bettman, and mary …...aquino, and levy 2007; winterich, mittal, and...

Documents